Signal processing for steerable sensor systems – Final report


  • Fredrik Näsström
  • Jonas Allvar
  • Staffan Cronström
  • Viktor Deleskog
  • Erika Bilock
  • Philip Engström
  • Fredrik Hemström
  • Gustaf Hendeby
  • Jörgen Karlholm
  • Roland Lindell
  • Sebastian Möller
  • Joakim Rydell
  • Per Skoglar
  • Karl-Göran Stenborg
  • Morgan Ulvklo
  • Joakim Wikström

Publish date: 2012-12-31

Report number: FOI-R--3570--SE

Pages: 46

Written in: Swedish


  • detection
  • tracking
  • collaboration
  • fusion
  • sensor management
  • situa-tional awareness


This report presents the work carried out within the project Signal processing for steerable sensor systems (S3). The aim of the project has been to demonstrate the feasibility and benefits of optimized automatic data collection towards requested reconnaissance, surveillance and security missions. A scenario that has been studied is how sensor data from mobile land vehicles and UAVs can be combined to provide improved local situation awareness around your own convoy to reduce the risk of, e.g., ambush. The development of platforms with steerable sensor systems has been very strong in recent years so some examples of military platforms with such systems are presented in this report. The report presents the work that has been undertaken in the automatic detection and tracking of people with IR sensors on ground vehicles and UAVs. The further developed detection algorithm is based on machine learning and can also be trained to detect other objects such as vehicles, ships, containers etc. Collaboration between sensors on UAVs and vehicles has been studied in the pro-ject. For instance, by merging information from sensors on a UAV and a ground vehicle, the accuracy in positioning of a detected target can be significantly im-proved. Fusion between terrain data and sensor data to improve accuracy in the positioning of both the sensor platform and detected targets has also been studied in the project. Work has begun on researching image analysis methods to automatically identify regions of interest in IR imagery, e.g., forests, buildings etc., where threats usually arise. With this functionality, search patterns and detection algorithms can be opti-mized depending on the scene content.